OrnitWith the Alu insertion where you are switching splice forms, one would expect that expressing a Iglllb splice form of FGFR2 in mesenchyme would be catastrophic to the developing embryo, in that it may activate an autocrine loop within mesenchymal tissue, leading to activation of the receptor. Why don't you see that?
Wilkie: We would expect all these FGFR mutations to be catastrophic. We know that FGFR1 and FGFR2 are essential early in embryogenesis and the knockout mutations are lethal (Deng et al 1994, Yamaguchi et al 1994, Arman et al 1998). We don't understand why any of these activating mutations are compatible with life at all. I suspect the answer to your question is another side of that same coin. Clearly, what we have demonstrated is ectopic IgIIIb isoform expression in a fibroblast cell line from a child. This is completely different from saying that we know that the same thing is happening in the mesenchyme in very early limb buds. There is no way we can demonstrate that in the human, so at the moment this is only a hypothesis. But I think it's a reasonable working hypothesis.
Newman: A recent paper has shown that fibroblasts from Apert's patients overexpress transforming growth factor (TGF)^ (Locci et al 1999). How do you see that in this context?
Wilkie: The osteoblasts from these patients show alterations in the expression of many different molecules, not just TGF^, but also extracellular matrix and osteogenesis markers, like type I collagen, glycosaminoglycans and alkaline phosphatase (Lomri et al 1998, Fragale et al 1999, Locci et al 1999). Some of the findings (for alkaline phosphatase, for example) have been inconsistent, and it is hard to dissect out what is primary, and what are secondary effects of alterations in the growth characteristics of the cell. At the moment, what has happened is that people have picked out a few candidates and looked at their expression. They have had good reasons to choose those candidates, but we don't know the context to understand the significance of the results in relation to overall changes of expression. This is an obvious situation where microarray analysis, where you can analyse in parallel a large number of different transcripts, is going to be helpful to give an idea of the real pattern (Iyer et al 1999).
Beresford: Can I push you a little bit on the ligands in this setting? I got completely lost when you were talking about this strange finding that the plasmon resonance suggests that it is the binding to FGF2 that is altered. This immediately creates problems, because we don't really know how that gets out of the cell. But then you went on to say that in any case FGF2 knockouts are normal. Why did you make that point? If the phenotype is hypothetically due to tighter binding of or reluctance to release a bound ligand, which we suspect might be FGF2, why would the fact that the FGF2 knockout doesn't have a phenotype be informative?
Wilkie: Presumably you need an FGF in the suture to get signalling. If FGF2 was the only important FGF in the suture, you would expect there to be a phenotype with the knockout. Obviously, there could be redundancy — in fact, this is very likely. One explanation of the knockout result could therefore be that there is another FGF that can take on the role of FGF2 in that situation. I was really trying to make the point that there are various ligands expressed in cranial sutures, but we don't know what they're doing and which are the important ones. The pattern of enhanced FGF binding that we see with these mutations should enable us to make some predictions about what properties those ligands should have. For example, FGF4 wasn't found to have any greater affinity for the Apert mutations than for wild-type (Anderson et al 1998). It is therefore hard to make an argument that FGF4 is mediating the craniosynostosis, which is consistent with the low expression level of FGF4 in the sagittal suture (Kim et al 1998).
Ornit^: We have made the equivalent Apert syndrome mutation in the context of FGFR1. This has an increased affinity for FGF2 as well. This makes sense, because the mutation is in the linker domain.
As far as the FGF2 knockout mouse is concerned, it does have several phenotypes. There is not necessarily a suture phenotype, although I am not sure how closely this was looked at, but there are effects in migration of neurons in the cortex, and a vascular smooth muscle phenotype. An intriguing observation has been made by Kim et al (1998), showing that FGF9 is also expressed at high levels in the suture. In your assays, have you looked at FGF9 in the surface plasmon resonance assay?
Wilkie: I don't know how this fits in with your own studies on ligand binding, but in fact, FGF9 wasn't found to bind significantly to the constructs.
OrnitIt should bind.
Wilkie: Have you looked at FGF9 binding to your proline to arginine mutation in FGFR1?
OrnitNot yet. I can add that we have also made Fgf9 knockout mice. These mice die at birth, but there is no obvious suture defect.
Kingslej: In the crystal structure that you showed, you pointed to an arginine: is that the arginine residue involved in thanatophoric dysplasia?
Wilkie: Yes. The mutation is Arg248Cys in FGFR3 (Tavormina et al 1995).
Kingslej: The mutations that are affecting affinities or off-rates for the other ligands would actually be the proline residue at position 250 in FGFR3. Can you look at the crystal structure and see why the proline to arginine substitution at that position might affect affinity? Then, on the ligand side, if you look across the various FGF ligand family members, can you also predict which ones are likely to be affected?
Wilkie: I learnt an interesting thing about the black art of crystallography from this. Although the paper is published in Cell (Plotnikov et al 1999), apparently authors are allowed to hold on to the co-ordinates for nine months after publication. Thus the coordinates are not in the public domain. All that our crystallographer Yvonne Jones can go on is the stereo diagram. There are apparently now fancy programs that can be used to reconstruct a 3D structure from stereo diagrams! Basically, what you're asking is a very apposite question, but without the coordinates it is not possible to answer it. It is too subtle a thing to look at unless you know exactly where those residues are.
Ornit^: I would like to make one comment about the crystal structure. The crystal that was made only contains a fragment of the FGF receptor: it is missing some critical sequence between Ig domains 1 and 2 at the 5' end of the N-terminus of the receptor. Although Ig domain 1 probably has very little to do with ligand binding, we have shown in a number of different experiments that the region between Ig domains 1 and 2 probably constitutes a second ligand binding site, in particular for FGF2, and may be a primary binding site for other ligands such FGF9 in different receptors. I think the crystal structure is therefore by no means the whole story, and they haven't really addressed how dimerization occurs. The other issue is that the receptor dimer that they show in the crystal structure is based on an FGF and FGFR in an asymmetric unit, and not an actual crystallization dimer of the receptors. Some of the interactions that they are showing may not be what's really happening. One final point is that the crystal was made in the absence of heparin, which we know is also a critical factor for binding and receptor activation.
Wilkie: They seem most confident about the ligand-receptor interaction.
Wilkins: It would be nice to study some of these things developmentally. Is anybody engineering mice with a conditional ectopic expression of these mutant forms?
Wilkie: Certainly. The bone dysplasias have led the way: there are now quite a few mouse models of these (Xu et al 1999, Li et al 1999, Wang etal 1999, Chen et al 1999, Segev et al 2000). The craniosynostosis models have been coming more slowly, and I don't think any have yet been published. The proline to arginine mutation in FGFR1 has now been made (Zhou et al 2000) and there are others coming on line.
"Newman: I would like to press you a bit on the TGF^ story. It seems to me that no matter how indirect the receptor's effect is on TGF^, if osteoblasts and precartilage mesenchymal cells are producing more TGF^ and consequently producing more fibronectin and so on, then that would be explanatory of many of the syndromes that you see. It seems to me that it doesn't really matter exactly what mechanism the receptor is using to do it, if it's causing an increase in molecules that promote mesenchymal condensation, that would be your explanation.
Wilkie: I can't add any further comment.
Kronenberg: I like your hypothesis about the syndactyly and the altered splicing. This makes this mutation qualitatively different from the others, and not just a quantitative part of a continuum. In this context, why do some of these mutations result in craniosynostosis and others result in achondroplasia, but never both? Or are they ever both?
Wilkie: Thanatophoric dysplasia combines both craniosynostosis and short-limbed bone dysplasia. It tends to be classified as a short-limbed bone dysplasia, because craniosynostosis is a clinical problem that usually presents in childhood rather than at birth, and thanatophoric dysplasia is neonatally lethal. In this disorder the two phenotypes can occur together. But I would make the general comment that there are some extremely subtle genotype—phenotype correlations that we still don't understand. A question I wanted to ask David Ornitz yesterday, which I will raise now, is that one of the FGFR3 mutations (Ala391Glu) causes a combination of both a Crouzon-like syndrome with craniosynostosis, and a skin disorder called acanthosis nigricans (Meyers et al 1995): do you know of any work that has elucidated the mechanism of that particular mutation?
Ornit^: I am not aware of any biochemical work on that mutation. FGFR3 is expressed in skin, in epidermis and possibly dermis. It is also expressed in a chondrocyte layer underneath the cranial sutures. At least from patterns of expression there is some precedent for a phenotype in these tissues. It's possible that some of these mutations may affect ligand binding specificity, perhaps in more subtle ways than specificity induced by alternative splicing, but they also could affect heterodimerization between FGF receptors, which may give completely novel phenotypes.
Wilkie: That mutation is a particularly good candidate for affecting FGFR heterodimerization (Shi et al 1993).
Mundlos: Can you tell us a little more what MSX is doing? Is it controlling bone growth?
Wilkie: I haven't done any of this work myself: it is based on work in the mouse, from Rob Maxson's lab in collaboration with Richard Maas. He suggests that MSX is involved in controlling both proliferation and differentiation. He proposes that a reduced dose of MSX2 reduces proliferation, and so there is a smaller number of cells that are able to commit to differentiation (Liu et al 1999, Dodig et al 1999). The way I would interpret our own findings, is that quite unequivocally the reduced MSX2 dosage has reduced the amount of differentiation. The actual size of the skull is normal — there's no effect on the head circumference — so it's impossible to comment on the effect on proliferation based on the human phenotype.
Mundlos: So it is an ongoing process.
Wilkie: Yes, it's ongoing in the sense that these parietal foramina persist throughout life but slowly lessen in size.
"Newman: I have a hypothesis that I would like to present that relates to the syndactyly seen in Apert syndrome. This is based on results from chick embryology. We became interested in the FGF receptors because we were looking at a culture system where we would get foci of precartilage condensation forming in culture. This would expand and eventually cover the whole culture unless we added ectoderm or FGFs, in which case the condensations and nodules that formed from them would remain confined. We had earlier shown that TGF^ was responsible for producing fibronectin in these nodules (Leonard et al 1991, Downie & Newman 1995). Because TGF^ is positively auto-regulatory, we would get a natural expansion. But with ectoderm, which is a source of various FGFs, we would get perinodular inhibition. If you look at the development of the limb, in the pre-condensed mesenchyme the main FGF receptor present is FGFR1. In precartilage condensations, the main receptor present is FGFR2. When cartilage is differentiated, FGFR3 is present (Szebenyi et al 1995). Therefore the relevant FGFR in precartilage condensations is FGFR2. Now if a source of FGFs causes a confinement of the condensations, then activating FGFR2 may be causing the release of the putative lateral inhibitor for a Turing-type reaction—diffusion system. To reiterate, this is a system in which there is positive auto-regulation of an activator, which also induces the release of a lateral inhibitor of its own activity. If that inhibitory effect was abrogated by a mutation in FGFR2, then without the inhibition you would get expansion of condensations and syndactyly rather than individual elements. Let's see if this relates to these actual mutations. Since we are dealing with mesenchymal cells, you would expect to have the IgIIIc type of FGFR2. This is generally what's seen here. But if you have a mis-splicing so that you have the IgIIIb form, then you would have the wrong kind of signalling: you would have a receptor that wouldn't be appropriate here, and whatever signal transduction that might lead to these the inhibitory effects could be abrogated under these conditions. The other mutations that preserve the IgIIIc form are described as gain-of-function mutations, because they bind more strongly to FGF, but my understanding is that it's only FGF2 that binds more strongly to the mutated IgIIIc forms.
Wilkie: This is the point about the two classical Apert mutations showing opposite effects in the severity of craniofacial malformations versus the syndactyly. The experiments that were done on the IgIIIc form are in keeping with the severity of the craniofacial problems, but they're actually the reverse of what you see for the syndactyly. This suggests that this is not the mechanism ofthe syndactyly.
"Newman: The Iglllc is the normal one: when you get the Iglllb form that arises from the Alu insertions, do those patients exhibit syndactyly?
Wilkie: Yes, by definition, because that's why they have got Apert syndrome (Oldridge et al 1999). But what I was trying to get at is the question of why you get syndactyly in classical Apert syndrome. What we need to do is to close the loop by repeating the experiments on the IgIIIb form. What I would hope to find is that there is a ligand like FGF10 which binds with enhanced affinity to the IgIIIb form, but this time the enhancement in affinity is greater for the Pro253Arg mutation than it is for the Ser252Trp mutation. This would then explain the opposite effects on severity between the two mutations.
Ornit^: Why would you expect to have an Iglllb form of the receptor in the classical Apert mutation? It shouldn't affect splicing.
Wilkie: The site of the mutations (in the linker between IgII and IgIIIa, see Fig. 2) is such that they are present both in the normal IgIIIb isoform and the IgIIIc isoform.
Kronenberg: Do you know you don't affect splicing? Adjacent sequences can sometimes affect the binding of proteins that regulate splicing.
Wilkie: We have excluded this (Oldridge et al 1999).
"Newman: In any case, there are two possible ways of abrogating the normal signal that you would get from the FGFR2. One would be mis-splicing, so that you would have the IgIIIb form in the mesenchymal tissue rather than the IgIIIc form. Another way of doing it is to have enhanced affinity for a ligand that is not the normal ligand for this process. For example, the endoderm is putting out a whole variety of FGFs, including FGF2 and FGF4. If the receptor exhibited enhanced affinity for FGF2, for example, and that was not the ligand responsible for mediating this inhibitory effect, then you might have competition with the normal ligand. The crux of this hypothesis is that it's the expression in the condensations of the correct form of FGFR2, with the appropriate ligand, that is responsible for mediating an inhibitory effect on chondrogenesis. If that's abrogated, then you get expansion of the condensations and syndactyly.
Kingsley: What's the phenotype of a simple FGFR2 knockout?
Wilkie: This is an early embryonic lethal, owing to a defect shortly after implantation (Arman et al 1998). An independent construct with leaky expression is lethal at day 10—11, with absent limbs (Xu et al 1998).
Newman: In our culture system, we've done antisense oligos against FGFR2, and have got expanded condensations. It's essentially a knockout in the culture. When we remove FGFR2, instead of getting individual condensations we get the whole mass of the culture turning into a sheet of cartilage.
Bard: And if you take out the antisense oligos, what happens then?
Newman: We are electroporating these constructs in, so we can't easily take them away.
Kronenberg: Why do you see this effect in the culture? Is it because of proliferation?
Newman: We don't know. Normally, these cells are not really proliferating very much at all. TGFjS is causing matrix production, and fibronectin and cell surface adhesive proteins are causing the precartilage cells to condense. TGF^ will spread out from those sites because of its positive auto-regulation and diffusion, and tend to cause an expansion of these centres. If you have a lateral inhibitor of TGF^ the expansion will be restricted, and you will get spaced-out elements. This is why I found the report of Locci et al (1999), that Apert syndrome fibroblasts and osteoblasts have enhanced TGF^ so interesting. This is exactly what is predicted by this model: the cells are not responsive to whatever is causing TGF^ to limit itself.
Kronenberg: So in your oligo experiment, where you remove FGFR2, why do all the cells become chondrocytes?
Newman: There is no FGFR2 at these centres, so when the condensations meet with ectodermal products, such as FGFs, no inhibitor is moving out from those centres, and we therefore get an expansion of the condensations.
Mundlos: From a clinical perspective this would mean that individuals with syndactyly should have very large bones.
Newman: You would see hard tissue fusion.
Mundlos: Whereas in most syndactyly cases there are normal bones with different fusions, either bony fusions or soft tissue fusions. My understanding was that this is more commonly attributed to the effects of BMPs in apoptosis, rather than these anlage being enlarged. In this case we would expect a hand like in the noggin mutations, which is all bone and cartilage.
Newman: Some of the X-rays that I've seen in published reports have very misshapen bones.
Wilkie: In Apert syndrome, I wouldn't say the bones were increased in size, in proportion to the soft tissues. As regards the shape of the bone, you get both longitudinal fusions of the proximal and middle phalanges, and transverse fusions involving the distal phalanges of the middle three digits, so they're fused into one single bone (Fig. 1E). These fusions tend to reflect the morphology of the autopod as a whole. I would see them in terms of the actual mesenchymal condensations being primarily abnormal, rather than as a secondary process of failure of apoptosis.
Ornit^: But there is no change in the length of the bones when they are fused?
Wilkie: If anything they're a little bit shorter than normal.
Kingsley: These were the hand phenotypes: in the clinical picture you showed it looked like the foot phenotypes are more confined to soft tissues.
Wilkie: In the feet there are certainly no transverse fusions. As you know it's often quite difficult to see the foot phalanges on X-ray, so I wouldn't want to comment on longitudinal fusions. In fact, the hallux of the foot is often quite strikingly abnormal with pre-axial polydactlyly involving the first metatarsal bone (Cohen & Kreiborg 1995).
OrnitWould you like to comment on why the bone fusions are restricted to the digits and not to more proximal bones?
Wilkie: This is not strictly true: there are other fusions that can occur. One that is quite frequent with FGFR2 mutations — in fact, I find it a useful clinical clue as to whether to look for mutations — is elbow fusion. Humero—radial or humero— ulnar synostosis are both described (Cohen 1993, Cohen & Kreiborg 1993). You can also get longitudinal vertebral fusions, particularly of the cervical vertebrae (Thompson et al 1996).
Morriss-Kay: It is also potentially informative that the digital fusions are progressively severe from proximal to distal, which could suggest that whatever the mechanism is, the longer the cells in the progress zone are exposed to it, the more they are prone to the loss of interdigital tissue. In that sense, what the digits tell you is that you would expect there to be the mildest effect on the most proximal elements of the limb.
Regarding the IgIIIb and IgIIIc isoform expression patterns, Sachiko Iseki in my lab has carried out isoform-specific in situ hybridization for Fgfr2 on mouse limb buds. Although in general the IgIIIb isoform is expressed in epithelia and IgIIIc isoform in mesenchyme, when you look at the development of the hand plate, it's actually not as clear-cut as that. Around the condensations and in the interdigital mesenchyme there is strong expression of the IgIIIc splice variant, but there is also some expression of IgIIIb. It could be that there is an interaction between the two splice variants, or that, when co-expressed, they're both working towards the same end. It is important not to have hypotheses concerning isoform-specific functions on the assumption that their expression patterns show a simple epithelial—mesenchymal reciprocity.
Newman: But even if there's a balance of the two isoforms, if there was a mutation that shifted that balance, it could have the hypothesized effect.
Morriss-Kay: Certainly, one would expect that to have an effect. This is what was recently suggested on the basis of altered isoform expression in Apert syndrome (Oldridge et al 1999).
Was this article helpful?